AI BriefingWorkday cuts AI agent prototype time from 5 months to under 1 day using LangGraph and LangSmith
·2026.11.11 01:00
Key point
A team of four engineers built the platform on Workday's infrastructure, which has since grown to roughly 30 members.
Details
- Workday built an internal developer platform for AI agents using LangGraph for orchestration and LangSmith for tracing, observability, and evals.
- Prototype development time dropped from 4 to 5 months to under 1 day, with authentication and security requirements baked into the staging environment.
- The platform team grew from 4 to roughly 30 members following the successful deployment of the system.
- Workday created a Python SDK called "Workday LangChain" that wraps LangChain primitives and routes them through internal services like an LLM gateway and RAG retrieval service.
- The platform includes an Agent System of Record, a governance layer that controls which APIs agents are allowed to access.
- LangGraph's graph-based model allows engineers to build simple ReAct loops to complex deterministic workflows without platform constraints on agent design.
- All agent servers run with built-in support for A2A and MCP protocols, ensuring immediate compliance without additional backend work.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Arcee releases nac, an open-source agent harness for long-running tasks